[PATCH] D141011: AMDGPU: Set scratch_en if there is dynamic stack but no fixed stack

[Matt Arsenault via Phabricator via llvm-commits]

arsenm created this revision.
arsenm added reviewers: AMDGPU, rampitec.
Herald added subscribers: kosarev, foad, kerbowa, hiraditya, tpr, dstuttard, yaxunl, jvesely, kzhuravl.
Herald added a project: All.
arsenm requested review of this revision.
Herald added a subscriber: wdng.
Herald added a project: LLVM.

https://reviews.llvm.org/D141011

Files:
  llvm/lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp
  llvm/test/CodeGen/AMDGPU/enable-scratch-only-dynamic-stack.ll


Index: llvm/test/CodeGen/AMDGPU/enable-scratch-only-dynamic-stack.ll
===================================================================
--- /dev/null
+++ llvm/test/CodeGen/AMDGPU/enable-scratch-only-dynamic-stack.ll
@@ -0,0 +1,20 @@
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 --amdhsa-code-object-version=5 < %s | FileCheck -check-prefixes=GCN,COV5 %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 --amdhsa-code-object-version=4 < %s | FileCheck -check-prefixes=GCN,COV4 %s
+
+ at gv.fptr0 = external hidden unnamed_addr addrspace(4) constant ptr, align 4
+
+; No stack objects, only indirect call has to enable scrathch
+; GCN-LABEL: test_indirect_call:
+
+; COV5: .amdhsa_private_segment_fixed_size 0{{$}}
+; COV4: .amdhsa_private_segment_fixed_size 16384{{$}}
+
+; GCN: .amdhsa_user_sgpr_private_segment_buffer 1
+
+; COV5: .amdhsa_uses_dynamic_stack 1
+; GCN: .amdhsa_system_sgpr_private_segment_wavefront_offset 1
+define amdgpu_kernel void @test_indirect_call() {
+  %fptr = load ptr, ptr addrspace(4) @gv.fptr0
+  call void %fptr()
+  ret void
+}
Index: llvm/lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp
===================================================================
--- llvm/lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp
+++ llvm/lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp
@@ -908,7 +908,8 @@
   // anything to disable it if we know the stack isn't used here. We may still
   // have emitted code reading it to initialize scratch, but if that's unused
   // reading garbage should be OK.
-  const bool EnablePrivateSegment = ProgInfo.ScratchBlocks > 0;
+  const bool EnablePrivateSegment =
+      ProgInfo.ScratchBlocks > 0 || ProgInfo.DynamicCallStack;
   ProgInfo.ComputePGMRSrc2 =
       S_00B84C_SCRATCH_EN(EnablePrivateSegment) |
       S_00B84C_USER_SGPR(MFI->getNumUserSGPRs()) |


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D141011.486375.patch
Type: text/x-patch
Size: 1796 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20230104/a29de76f/attachment.bin>